In the Tools section, V.O. experts review online tools for IE-aware innovators. In this edition, we look at ChatGPT, an experimental AI-driven chatbot based on a so-called large language model. The tool has already received much attention for its ability to understand human prompts and generate all kinds of texts. Since texts play an important role in the world of IE, the question is what ChatGPT has to offer in that regard.
On ChatGPT’s opening screen, we immediately notice a list of “limitations,” such as “May occasionally generate inaccurate information.” Clicking through to an FAQ section, we find even more warnings: “ChatGPT will sometimes fabricate facts or hallucinate output”, “please do not share sensitive information in your conversations” and “your conversations may be reviewed by our AI trainers”.
Engaging with AI on IP
With all this in mind, we begin by asking ChatGPT the question, “Can you give me some advice on intellectual property?” The AI happily replies, “Sure! Here’s some general advice on intellectual property,” followed by a list of seven items of explanation, from “Understand the different types of intellectual property” to “Work with a professional” (our favorite).
What follows is a detailed discussion of IP, which feels distinctly human and demonstrates surprisingly thorough knowledge that is highly relevant to our questions. Moreover, presented in a way that is easy to understand. There is no denying that ChatGPT, as a prominent example of a new wave of “generative” AI models, represents a great leap forward in the evolution of AI.
Helping
ChatGPT itself indicates in our conversation that it can help with patent and trademark searches, drafting and analysis. While AI tools can indeed be useful, the key word remains “help”. ChatGPT clearly does not (yet) work on the same level as a human patent or trademark attorney. For now, ChatGPT does not know and understand important, current legislation and case law, and it is questionable whether it will ever develop a sense of justice.
Importantly, ideas shared with ChatGPT could be considered public. If so, that would eliminate novelty for a future patent application.
In any case, ChatGPT itself acknowledges that it sometimes provides incorrect information. This is extremely risky when working in a legal context. We ourselves encountered a so-called “hallucination” when ChatGPT assured us that V.O. has an office in Spain. This is factually incorrect, but perhaps ChatGPT accidentally generated an excellent idea….